GitHub - firecrawl/firecrawl: Knowledge infrastructure to make your agents smarter. Search, scrape and access 550+ data providers through one API.

Open original

Firecrawl is an open-source toolkit that converts websites into clean markdown, with a local binary, a hosted cloud service and a self-hostable server. The repo README explains how to install the local version, plug it into AI tools as an MCP server, and call the same API from existing SDKs.

  • Firecrawl Local runs as one binary from crates.io, scraping, crawling and mapping sites without a database, queue or container, and it works with Claude Code, Cursor, Windsurf and similar tools via MCP.
  • Pages that need JavaScript render in a Chrome or Chromium already on your machine, and the local server answers the Firecrawl v2 API on port 3002 so existing SDK code runs unchanged.
  • Firecrawl Cloud adds search, structured JSON, interact and agent features, plus access to 550+ data providers through Alexandria, with a free tier of 1,000 credits a month.

Firecrawl's repo ships three ways to use the product. Firecrawl Local is a single binary you install with `cargo install firecrawl-local` (Rust and a C compiler required). It turns pages into markdown on your own machine, with no database, queue, or container, and it can register itself as an MCP server across Claude Code, Claude Desktop, Cursor, Windsurf, Gemini CLI, and Codex via `firecrawl-local init`. Running `firecrawl-local serve` exposes a Firecrawl v2-compatible API on port 3002, so existing SDK code works if you just change the API URL to localhost. Self-hosting deploys the full server from `apps/api` with its own queues, workers, and databases, and the hosted Firecrawl Cloud service covers the rest. The split between Local and Cloud is where the practical decisions sit. Local handles scrape, crawl, map, and batch scrape, plus PDF and Office parsing, and it renders JavaScript-heavy pages through whatever Chrome or Chromium is already installed. Without a browser, static pages still work and pages that needed one return a warning. A single-page scrape that fails locally falls back to Cloud and is tagged `"source": "firecrawl-cloud"` in the result; setting `FIRECRAWL_CLOUD=off` disables that fallback. Search, the Alexandria data catalog (550+ providers), Interact, Agent, and Monitor are Cloud-only. Cloud also adds structured JSON output, summaries, and screenshots. Keyless use gets 1,000 free credits a month without an account, and `firecrawl-local login` ties Local's Cloud calls to a real account. The README's examples are mostly code snippets for search, scrape, interact (driving pages with natural-language prompts like "Click the first result"), and agent (a prompt-only call that returns an answer and its sources). Cloud claims include 96% web coverage, a P95 latency of 3.4s, and LLM-oriented output to reduce token use. Those are the vendor's own figures, so treat them as marketing until you test them on your targets. The article is also heavy on repeated star-request and signup nudges, which you can skip. For a read-it-later pick, the useful bit is the Local-versus-Cloud boundary and the MCP setup, since that determines whether you can run the whole thing without paying anyone.

Ask about this article